CN103366178B - A kind of method and apparatus for being used to carry out target image color classification - Google Patents
A kind of method and apparatus for being used to carry out target image color classification Download PDFInfo
- Publication number
- CN103366178B CN103366178B CN201210091353.8A CN201210091353A CN103366178B CN 103366178 B CN103366178 B CN 103366178B CN 201210091353 A CN201210091353 A CN 201210091353A CN 103366178 B CN103366178 B CN 103366178B
- Authority
- CN
- China
- Prior art keywords
- color
- target image
- classification
- hsv
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Abstract
It is an object of the invention to provide a kind of method and apparatus for being used to carry out target image color classification.Specifically, the present invention by hsv color space by being divided into multiple color sub-spaces, according to HSV assignment of the target image under the hsv color space, determine the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image, to determine the color classification of the partly or entirely pixel, according to the color classification of the partly or entirely pixel, determine the color classification of the target image, solve the problems, such as that existing method is not applied for various different application scenes, improve the accuracy of method for sorting colors, the usage experience of the further method for sorting colors efficiency and user for improving target image.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of hsv color space that is based on to carry out color to image
The technology of classification.
Background technology
With the rapid development of electronic technology and digital processing technology, the image electronic equipment such as digital camera, scanner
The raising of popularity rate, picture material just increase with surprising rapidity.How institute is effectively retrieved from the view data of magnanimity
The problem of image needed is one very crucial.Such as user want retrieve a photo similar to Beethovan's appearance, usual one
As text search engine for graph image, almost feel simply helpless.Therefore, effective retrieval to image becomes us
Obtain the key issue of image information.
Two methods of current image retrieval (Image Retrieval) generally use:First, the figure based on text index
As retrieval, second, CBIR.The system of image retrieval based on text index, all it is first to be described with text
The semanteme of image, then using ripe text searching algorithms, such as Page-Rank side on the basis of these image texts mark
Method, probabilistic method, location method, method of abstracting, classification or clustering method, part-of-speech tagging method etc., scheme for user's search is expected
Picture.But with the increase of image data base scale, more than several hundred million, the problem of image retrieval based on text index is present, is just
Highlight, manually image is labeled wastes time and energy very much first, and secondly artificial mark has subjectivity and uncertainty,
Such as same piece image, the mark that different people provides may be entirely different, and this make it that the inquiry for responding user exactly is non-
It is often difficult.CBIR, it is not necessary to the participation of user, and utilize image itself feature, as color, texture,
The features such as shape are retrieved, and have a stronger objectivity.Generally, can be used with the feature of all images in abstract image storehouse
The process of family retrieval is typically to provide a sample image, and system extracts the individual features of the sample image, then same database
In all features of object that are retrieved be compared, and the image similar to sample feature is returned into user.
Color characteristic is the visual signature being most widely used in CBIR, main reason is that color
It is often highly dependent with the object included in image or scene.In addition, compared with other visual signatures, color characteristic is to figure
As the dependence at the size of itself, direction, visual angle is smaller, so as to have higher robustness.Conventional color characteristic represents at present
Method has the expression side of the color characteristics such as color histogram, color moment, color set, color convergence vector and color correlogram
Method.At present to the color classification of target image mainly using being converted into target image by image processing techniques to use above-mentioned face
The statistic of color characteristic, color space is such as divided into several small color intervals, then fallen by calculating color each
Pixel quantity in minizone can obtain the color rarity of target image.Current method for sorting colors can not the property taken into account
Energy and efficiency requirements, the number such as color minizone is more, stronger to the resolution capability of color, but can increase computation burden,
It is unfavorable for applying and index is established in Large Graph is as library searching;The number of color minizone is few, although computing cost is smaller,
The resolution capability of color is relatively low, is unfavorable for applying and can't stand the application to associated picture mistakes and omissions.Take into account performance and efficiency
One core key technology of method for sorting colors is that solve the problems, such as color space Subspace partition, empty to obtain multiple color
Between, it is then determined that the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image, to determine
The color classification of the partly or entirely pixel.But current target image method for sorting colors does not solve color space
Space partition problem, and then cause current method for sorting colors to be not applied for a variety of application scenarios;Current face
Colour sorting method does not distinguish to the band of position of pixel in target image yet, and each pixel in target image is in color classification
In shared weight it is the same, and then cause current method for sorting colors to carry out color classification to target image exactly,
If piece image is left black right white, another piece image is right black left white, and wherein black-white colors respectively account for 50%, the two width color
Histogram is the same, but this two images difference is obviously.Obviously, it is empty not solve color space for existing method
Between partition problem, and then cause existing method to be not applied for a variety of application scenarios;Existing method is to target figure
The band of position of pixel does not distinguish yet as in, and then causes the accuracy of method for sorting colors not high yet, reduces target figure
As carrying out the accuracy of color classification, while reduce user experience.
Therefore, for above-mentioned both sides content, the method for target image color classification can how be made suitable for various
Different application scene, and can improve accuracy the problem of turning into those skilled in the art's urgent need to resolve of method for sorting colors, so
And currently there has been no the solution for this problem.
The content of the invention
It is an object of the invention to provide a kind of method and apparatus for being used to carry out target image color classification.
According to another aspect of the present invention, there is provided a kind of to be used to carry out color to target image by computer implemented
The method of classification, this method comprise the following steps:
X carries out Subspace partition processing to hsv color space, to obtain multiple color sub-spaces, wherein, each color
Space corresponds to a kind of color;
Wherein, this method also includes:
A obtains the target image for intending carrying out color classification;
HSV assignment of the b according to the target image under the hsv color space, determines part in the target image
Or the mapping relations of whole pixels and the multiple color sub-spaces, to determine the color classification of the partly or entirely pixel;
C determines the color classification of the target image according to the color classification of the partly or entirely pixel.
According to an aspect of the present invention, a kind of sorting device for being used to carry out target image color classification is additionally provided,
The equipment includes:
Space divides device, empty to obtain multiple color for carrying out Subspace partition processing to hsv color space
Between, wherein, each color sub-spaces correspond to a kind of color;
Wherein, the equipment also includes:
Image acquiring device, for obtaining the target image for intending carrying out color classification;
Space reflection device, for the HSV assignment according to the target image under the hsv color space, determine institute
The mapping relations of part or all of pixel and the multiple color sub-spaces in target image are stated, it is described part or all of to determine
The color classification of pixel;
Color determining device, for the color classification according to the partly or entirely pixel, determine the target image
Color classification.
In accordance with a further aspect of the present invention, a kind of browser for being used to carry out target image color classification is additionally provided.
In accordance with a further aspect of the present invention, a kind of search for being used to carry out target image color classification is additionally provided to draw
Hold up.
Compared with prior art, the present invention by hsv color space by being divided into multiple color sub-spaces, wherein supporting logical
The method for crossing machine learning carries out Subspace partition to color space, according to the target image under the hsv color space
HSV assignment, the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image are determined, with true
The color classification of the fixed partly or entirely pixel, according to the color classification of the partly or entirely pixel, determines the target
The color classification of image, wherein supporting with reference to the partly or entirely band of position of the pixel in the target image, weighting
The color classification of the target image is determined, solves the problems, such as that existing method is not applied for various different application scenes, is improved
The accuracy of method for sorting colors, the usage experience of the further method for sorting colors efficiency and user for improving target image.
Brief description of the drawings
By reading the detailed description made to non-limiting example made with reference to the following drawings, of the invention is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 shows to be illustrated according to the sorting device for being used to carry out target image color classification of one aspect of the invention
Figure;
Fig. 2 shows the sorting device for being used to carry out target image color classification in accordance with a preferred embodiment of the present invention
Schematic diagram;
Fig. 3 show according to a further aspect of the present invention by it is computer implemented be used for target image carry out color classification
Method flow diagram;
Fig. 4 show in accordance with a preferred embodiment of the present invention by it is computer implemented be used for target image carry out color
The method flow diagram of classification.
Same or analogous reference represents same or analogous part in accompanying drawing.
Embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings.
Fig. 1 shows to be illustrated according to the sorting device 1 for being used to carry out target image color classification of one aspect of the invention
Figure.Here, sorting device 1 includes but is not limited to individual host, minicomputer, large scale computer, multiple main frames collection, network host, single
The webserver, multiple webserver collection, and the cloud that multiple servers are formed.Here, cloud is by based on cloud computing (Cloud
Computing a large amount of computers or the webserver) are formed, wherein, cloud computing is one kind of Distributed Calculation, by a group pine
Dissipate a super virtual computer of the computer collection composition of coupling.
As shown in figure 1, sorting device 1 includes space division device 101, image acquiring device 102, space reflection device
103rd, color determining device 104.
Specifically, space division device 101 carries out Subspace partition processing to hsv color space, to obtain multiple colors
Subspace, wherein, each color sub-spaces correspond to a kind of color.
Wherein, HSV (Hue, Saturation, the Value) color space is three base attributes according to color:Color
Phase (Hue), saturation degree (Saturation) and lightness (Value) determine a kind of method of color.The model in hsv color space
Corresponding to a conical subset in cylindrical-coordinate system, the suitable form and aspect of rounded bottom surface of cone, saturation degree is then from the center of circle to side
Edge increase, lightness are then successively decreased from base to the vertex of a cone, and H, S, V span is respectively 0≤V≤1,0≤S≤1,0≤H < 360.
The top surface of circular cone corresponds to V=1, represents that color is brighter, and color H around the anglec of rotation of V axles by giving.Red corresponds to 0 ° of angle,
Green corresponds to 120 ° of angle, and blueness corresponds to 240 ° of angle.In hsv color model, each color and its complementary color phase
Poor 180 °.Saturation degree S values, so the radius of circular cone top surface is 1, the apex of circular cone, V=0, represent black from 0 to 1, circle
S=0, V=1 at the end face center of cone, white is represented, the summit from the end face center of circular cone to circular cone represents brightness gradually dark ash
Color, the i.e. grey with different gray scales.Other conventional color spaces also have RGB color, HSI color spaces, HSL colors
Space, HSB color spaces, YUV color spaces, Lab color spaces, XYZ color space, Ycc color spaces, CMYK color space
Deng.Relative to other color spaces, subjective understanding of the hsv color space closer to people to color, meet visual face
Color continuity, distribution of color is substantially continuous, is easy to the division of subspace.
For example, in hsv color space, V axles are divided equally into by space division device 101 with the space plane vertical with V axles
10 deciles, will the hsv color space be divided into 10 sub-spaces, if H values are 0-360, S values are 0-1, and V values are
0-0.1 color space is hsv color space subspace, and the subspace is labeled as HSV [0-360,0-1,0-0.1].Institute
Stating color sub-spaces includes each color spatial dimension corresponding in each color space subspace, such as red in color
Curve in space subspace HSV [0-360,0-1,0-0.1] is f1 (H, S) and f2 (H, S), is fallen in curve f1 (H, S) and f2
Pixel between (H, S) is i.e. labeled as red.
Skilled artisans will appreciate that the method for the above-mentioned subspace that 10 are divided into hsv color space is only to lift
Example, other are existing or are likely to occur from now on to the progress Subspace partition processing of hsv color space, empty to obtain multiple colors
Between method be such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
Image acquiring device 102 obtains the target image for intending carrying out color classification.
Wherein, the form of the target image includes but is not limited to:Bmp, jpg, tiff, gif, pcx, tga, exif,
Fpx, svg, psd, cdr, pcd, dxf, ufo, eps, ai, raw.
Wherein, the target image include being stored in local and or network computer or storage device in, wherein network meter
Calculation machine or storage device include cloud computing.
Wherein, the target image include by it is wired with or wirelessly from the computer or storage device
Obtain.
Such as image acquiring device 102, using HTTPs agreements, passes through number by internet from network picture database
The inquiry API provided according to storehouse, 1000 target images for intending carrying out color classification are obtained, such as " Garfield " image, it stores lattice
Formula is jpg picture.Wherein network picture database includes being stored on special server, and is stored in cloud computing platform
It is first-class.
Skilled artisans will appreciate that above by internet, using HTTPs agreements, obtained from network picture database
The method that plan carries out the target image of color classification is only for example, and other are existing or are likely to occur from now on for obtaining plan progress
The method of the target image of color classification is such as applicable to the present invention, should also be included within the scope of the present invention, and to draw
It is incorporated herein with mode.
Space reflection device 103, according to HSV assignment of the target image under the hsv color space, it is determined that described
The mapping relations of part or all of pixel and the multiple color sub-spaces in target image, to determine the part or all of picture
The color classification of element.
Wherein described HSV assignment refers to the HSV value corresponding to each pixel in target image, such as the picture in target image
HSV corresponding to plain x is H=0, S=0.5, V=0.05, is designated as HSV [0,0.5,0.05].
Wherein described mapping relations include the HSV value of each pixel in target image, and color space sky
Between divide, determine the subspace belonging to the pixel, then the color sub-spaces with the subspace are mapped.
For example, the HSV [0,0.5,0.05] of the pixel r in target image e, the hsv color space space vertical with V axles
V axles are divided equally into 10 deciles by plane, will the hsv color space be divided into sub-spaces of 10 sub-spaces, wherein one
Subspace belonging to individual is HSV [0-360,0-1,0-0.1], and space reflection device 103 determines that the pixel r belongs to subspace HSV
[0-360,0-1,0-0.1], space reflection device 103 are reflected pixel r HSV value and the color sub-spaces of the subspace
Penetrate, determine that pixel r falls between red curve f1 and f2, and then determine that pixel r belongs to red, space reflection device 103
Can by the way of the batch processing or distributed treatment mode to target image in partly or entirely pixel map, with
Determine the color classification of the partly or entirely pixel.
Skilled artisans will appreciate that the method that the HSV value of the pixel r in the above-mentioned e with target image determines color classification
It is only for example, other are existing or are likely to occur are assigned according to HSV of the target image under the hsv color space from now on
Value, the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image are determined, with described in determination
The method of the color classification of part or all of pixel is such as applicable to the present invention, should also be included within the scope of the present invention,
And it is incorporated herein by reference.
Color determining device 104, according to the color classification of the partly or entirely pixel, determine the face of the target image
Colour sorting.
For example, the color classification of the target image is determined according to the statistical information of color classification, the pixel of target image
The statistical information of color classification is:Red accounts for 60%, scarlet to account for 15%, orange red to account for 10%, and bright red accounts for 2%, and gold accounts for 2%, green
Color accounts for 3%, deep sky blue to account for 3%, the statistical information that color determining device 104 is classified according to the pixel color of target image, analysis
Red relevant colors account for 92%, and the color classification for determining the target image is red.
Skilled artisans will appreciate that the above-mentioned statistical information with by color classification determines the face of the target image
The method of colour sorting is only for example, and other are existing or are likely to occur color point according to the partly or entirely pixel from now on
Class, determine that the method for the color classification of the target image is such as applicable to the present invention, should also be included in the scope of the present invention
Within, and be incorporated herein by reference.
Preferably, it is continuous firing between each device of sorting device 1, specifically, space division device 11 continues
Subspace partition processing is carried out to hsv color space, to obtain multiple color sub-spaces;Image acquiring device 12, it is lasting to obtain
Intend carrying out the target image of color classification;Space reflection device 13, continue according to the target image in the hsv color space
Under HSV assignment, determine the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image, with
Determine the color classification of the partly or entirely pixel;Color determining device 14, continue according to the partly or entirely pixel
Color classification, determine the color classification of the target image.Constantly worked between above-mentioned each device, here, ability
Field technique personnel should be understood that " lasting " refer to above-mentioned each device respectively according to the mode of operation of setting or real-time adjustment require into
Row hsv color space carries out Subspace partition processing, the target image for obtaining plan progress color classification, determines the part or complete
The color classification of portion's pixel, and determine the color classification of the target image.
Preferably, space division device 101:
- by machine learning, Subspace partition processing is carried out to hsv color space, to obtain the multiple color sky
Between, wherein, each color sub-spaces correspond to a kind of color.
Wherein, the machine learning includes by computer simulation or realized the learning behavior of the mankind, and self-correction, from
It is dynamic to improve.
For example, space divides device 101 by machine learning, Subspace partition processing is carried out to hsv color space, to obtain
Red color subspace, it is as follows the step of machine learning:
Step 1:The positive negative sample of red color is first gathered, positive sample is that people's subjectivity is considered that red sample is added to
Positive sample set, negative sample are people's subjectivity not thinks it is that red sample adds negative sample set, and wherein color card is collected
It can be realized by successive ignition, the sample claims marked color card;
Step 2:In hsv color space, V axles are divided equally into 10 deciles with the space plane vertical with V axles, will be described
Hsv color space is divided into 10 sub-spaces;
Step 3:Initial red curve is first set in every sub-spaces, as red curve formula is:F1 (H, S)
=H+S^a1*b1+c1, f2 (H, S)=H+S^a2*b2+c2, wherein a, b, c are parameter, a, b, and c span is respectively:a
(1/10050), b (- 360360), c (- 360360), the mark that the HSV value of target image pixel falls between two curves are
Color;
Wherein initial parameter of curve is arranged to f1 (H, S)=H+S^0*0+0=0, f2 (H, S)=H+S^0*0+30=0;
Step 4:Marked color card is classified using original formulation, obtains ratio of the classification correctly with mistake;
Step 5:According to a, b, c spans, formula f1 and f2 are constantly updated according to step-length, often updates a step-length,
Marked color card is classified, correct and classification error a ratio is obtained, such as takes step-length constant for 5, a, b to c,
Obtain once new formula:F (H, S)=H+5=0 and f (H, S)=H+35=0, then using more new formula to marked color
Sample is classified, and obtains ratio of the classification correctly with mistake;
Step 6:After all testing all possible step size combination, one group of optimal parameter, i.e. classification error are obtained
One group of rate minimum;
Step 7:Using one group of optimal parameter, unmarked color card is classified, the color card of misclassification is added
Enter into corresponding marked color card, such as red sample is assigned in negative sample set, then the sample just adds
Into marked positive sample set, it is assigned to if not the color card of red in positive sample set, is then added to negative sample
In this set;
Step 8:According to the sample set of renewal, step 1 is continued executing with, is iterated, until classification accuracy reaches pre-
Determine threshold value, stop iteration, for example accuracy rate threshold value is 95%;
Step 9:Carry out the training of above-mentioned steps 1 to 8 respectively to each color.
It is skilled artisans will appreciate that above-mentioned by machine learning, to be carried out to hsv color space at Subspace partition
Reason, is only for example with obtaining the method for red color subspace, and other are existing or are likely to occur from now on by machine learning,
Subspace partition processing is carried out to hsv color space, the present invention is such as applicable to obtain the method for red color subspace,
It should be included within the scope of the present invention, and be incorporated herein by reference.
It is highly preferred that space divides device 101:
- according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space,
To obtain the multiple color sub-spaces corresponding with the color classification number, wherein, each color sub-spaces corresponding one
Kind color.
Wherein, the preset color classification number including the use of this method system or user according to application need or need
Ask, the color classification number pre-set.
Such as user is only interested in scarlet color, preset color classification number is 1, and space divides device 101 according to face
Colour sorting number performs above-mentioned steps 1 to 8, to obtain the corresponding color sub-spaces of scarlet color.
Skilled artisans will appreciate that be 1 above by preset color classification number, by machine learning, with obtain with
The method of the corresponding the multiple color sub-spaces of the color classification number is only for example, and other are existing or from now on may
Occur according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space, to obtain
The method for obtaining the multiple color sub-spaces corresponding with the color classification number is such as applicable to the present invention, should also include
Within the scope of the present invention, and it is incorporated herein by reference
Preferably, color determining device 104:
- according to the color classification and its respective weights of the partly or entirely pixel, weight and determine the target image
Color classification.
For example, color classification of the color determining device 104 according to each pixel, and the weight of each pixel, weighting determine institute
State the color classification of target image.Or for example, the face of the target image is determined by the statistical information weighting of color classification
Colour sorting, wherein red weights are 0.1, golden weights are 0.4, and green weights are 0.4, and blue weights are 0.1, target image
Pixel color classification statistical information be:Red accounts for 50%, and gold accounts for 20%, and green accounts for 20%, and blueness accounts for 10%, and color determines
The statistical information and weights that device 104 is classified according to the pixel color of target image, calculate the pixel of the target image after weighting
Statistical information be:Red accounts for 22.7%, and gold accounts for 36.3%, and green accounts for 36.3%, and blueness accounts for 4.5%, color determining device
104 according to the pixels statisticses information of the target image after weighting, and the color classification for determining the target image is golden+green.
Skilled artisans will appreciate that above-mentioned weight the determination target image to pass through the statistical information of color classification
The method of color classification be only for example, other are existing or are likely to occur color according to the partly or entirely pixel from now on
Classification and its respective weights, weighting determine that the method for the color classification of the target image is such as applicable to the present invention, should also wrapped
It is contained within the scope of the present invention, and is incorporated herein by reference.
It is highly preferred that color determining device 104:
- according to the color classification of the partly or entirely pixel, and with reference to the partly or entirely pixel in the target
The band of position in image, weighting determine the color classification of the target image.
Determined for example, being weighted by the band of position of the statistical information and pixel of color classification in the target image
The color classification of the target image, a width target image are square, and its picture material is an one-by-one inch photograph, and character image mainly collects
In in center section, target image draw is divided into 3 regions, respectively left, middle and right with the straight line perpendicular to base, wherein
Weights of the yellow in left, middle and right are respectively 0.1,0.8,0.1, and weights of the black in left, middle and right are respectively 0.1,0.8,0.1,
Weights of the blueness in left, middle and right are respectively 0.1,0.8,0.1, and the statistical information of the pixel color classification of target image is:Yellow
Account for 30%, black accounts for 30%, and blueness accounts for 40%, and wherein yellow accounts for 10%, 80%, 10% in left, middle and right respectively, black on a left side,
In, right to account for 10%, 80%, 10% respectively, blueness accounts for 40%, 20%, 40% in left, middle and right respectively, color determining device 104
The statistical information and the weights of present position classified according to the pixel color of target image, calculate the target image pixel after weighting
Statistical information be:Yellow accounts for 40.2%, and black accounts for 40.2%, and blueness accounts for 19.6%, after color determining device 104 is according to weighting
Target image pixels statisticses information, the color classification for determining the target image is yellow+black.
Skilled artisans will appreciate that above by the statistical information and pixel of color classification in the target image
In band of position weighting determine that the method for the color classification of the target image is only for example, other are existing or from now on may
There is the color classification according to the partly or entirely pixel, and with reference to the partly or entirely pixel in the target image
The band of position, weighting determines that the method for the color classification of the target image is such as applicable to the present invention, should also be included in this
Within invention protection domain, and it is incorporated herein by reference.
Preferably, sorting device 1 also includes the (not shown) of pretreatment unit 111, and pretreatment unit 111 is to the target figure
As being pre-processed, to obtain the pretreated target image.
Wherein, the pretreatment operation in the pretreatment unit 111 includes following at least any one:
- processing is compressed to the target image;
- color space mapping processing is carried out to the target image, to obtain the target image under hsv color space
HSV assignment;
- intercepting process is carried out to the target image.
Wherein, before the pretreatment is included in the color classification for determining image, the image is pre-processed, such as compression,
Interception or color space conversion etc., according to different application scenarios, certain conversion, such as target figure are carried out to target image
As color space be RGB color, it is necessary to be transformed into HSV image spaces.
Wherein, the compression processing includes pressing target image in the case where not losing excessive visual information
Contracting, as the lossless compression method of image has a Shannon-Fano codings, Huffman codings, the distance of swimming (Run-length) coding,
LZW (Lempel-Ziv-Welch) codings and arithmetic coding etc., the compression method of image have Karhunen-Loeve transformation coding and DCT to compile
Code etc..By compressing processing to reduce the workload that color of image is classified, color classification efficiency is improved.
Wherein, color space mapping processing is included the color space conversion of target image into hsv color space, such as
RGB color is converted into hsv color space.
Wherein, the intercepting process includes subregion or the gold visual zone by intercepting target image, to scheme
Exemplified by the trichotomy of picture, trichotomy by piece image it is vertical, be horizontally divided into three parts, i.e. piece image is divided into equal size 9 big
Small equal part, the joint position of straight line, i.e. middle position, for being laid out important visible elements, the region
Commonly referred to as gold visual zone, the golden section ratio of image is similar with trichotomy, where the joint of the straight line of golden section ratio
Position than trichotomy intersection point closer to picture centre.
For example, the color space mapping that the RGB color of target image is converted into hsv color space is handled, definition
Pixel s in target image is entered as (R, G, B) in rgb space, wherein R respectively, G, B represents the tax of red, green and blue respectively
Value, RGB span have normalized, and are the real number between 0 to 1, if max is R, the maximum in G, B, and min R, G, B
In reckling, pretreatment unit 111, color space mapping processing, RGB color are carried out to each pixel of the target image
The mapping treatment method in space to hsv color space is as follows:
Max=max (R, G, B)
Min=min (R, G, B)
IfR=max, H=(G-B)/(max-min)
IfG=max, H=2+ (B-R)/(max-min)
IfB=max, H=4+ (R-G)/(max-min)
H=H*60
IfH < 0, H=H+360
V=max (R, G, B)
S=(max-min)/max
It is highly preferred that space reflection device 103:
- HSV the assignment according to the pretreated target image under the hsv color space, determines the pre- place
The mapping relations of part or all of pixel and the multiple color sub-spaces in target image after reason, with determine the part or
The color classification of whole pixels.
For example, target image e RGB color is converted into by pretreatment unit 111 by color space mapping processing
Hsv color space, wherein the HSV [0,0.5,0.05] of the pixel r after processing, hsv color space is put down with the space vertical with V axles
V axles are divided equally into 10 deciles by face, will the hsv color space be divided into sub-spaces of 10 sub-spaces, one of them
Affiliated subspace is HSV [0-360,0-1,0-0.1], and space reflection device 103 determines that the pixel r belongs to subspace HSV [0-
360,0-1,0-0.1], space reflection device 103 is mapped pixel r HSV value and the color space of the subspace, it is determined that
Pixel r falls between red curve f1 and f2, and then determines that pixel r belongs to red, and space reflection device 103 can use
The mode of batch processing or the mode of distributed treatment are mapped part or all of pixel in target image, with described in determination
The color classification of part or all of pixel.
Skilled artisans will appreciate that mapped above by by target image e RGB color by color space
Processing is converted into hsv color space to determine that the method for the partly or entirely color classification of pixel is only for example, and other are existing
It is having or be likely to occur HSV assignment according to the pretreated target image under the hsv color space from now on, it is determined that
The mapping relations of part or all of pixel and the multiple color sub-spaces in the pretreated target image, to determine
The method for stating the color classification of part or all of pixel is such as applicable to the present invention, also should be included in the scope of the present invention with
It is interior, and be incorporated herein by reference.
Fig. 2 shows the sorting device for being used to carry out target image color classification in accordance with a preferred embodiment of the present invention
1 schematic diagram, wherein, sorting device 1 also includes image providing device 205.The preferred embodiment is retouched referring to Fig. 2
State:Specifically, space division device 201 carries out Subspace partition processing to hsv color space, empty to obtain multiple color
Between, wherein, each color sub-spaces correspond to a kind of color;Image acquiring device 202 is obtained and inputted with user by user equipment
The corresponding target image of search sequence;Space reflection device 203 is empty in the hsv color according to the target image
Between under HSV assignment, determine the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image,
To determine the color classification of the partly or entirely pixel;Color determining device 204 according to it is described partly or entirely pixel face
Colour sorting, determine the color classification of the target image;Image providing device 205 is by the target image and the target image
Color classification be supplied to the user equipment.Wherein, space division device 201, image acquiring device 202, space reflection dress
Put 203 and color determining device 204 it is same or similar with corresponding intrument shown in Fig. 1 respectively, therefore here is omitted, and by drawing
Mode is incorporated herein.Wherein, the user equipment includes PDA, mobile phone terminal, computer, the numeral with network savvy
Terminal etc..
Specifically, image acquiring device 202 obtains the institute corresponding with the search sequence that user is inputted by user equipment
State target image.For example, user uses mobile phone terminal, internet is connected to by WiFi modes, in network picture database
Text query information " blue Garfield " is inputted in cell-phone customer terminal, image acquiring device 202 uses HTTPs agreements, calls net
The API that network picture database provides, obtain the target image related to " blue Garfield ".
It is skilled artisans will appreciate that relative with the search sequence that user is inputted by mobile phone terminal above by obtaining
The method for the target image answered is only for example, other it is existing or be likely to occur from now on acquisition with user pass through user equipment
The method of the corresponding target image of the search sequence of input is such as applicable to the present invention, should also be included in present invention protection
Within scope, and it is incorporated herein by reference.
The color classification of the target image and the target image is supplied to the user to set by image providing device 205
It is standby.For example, user is connected to internet, in the cell phone customer of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in end, image acquiring device 202, using HTTPs agreements, calls network picture number
The inquiry API provided according to storehouse, obtains 1000 target images related to " blue Garfield ", and its storage format is jpg figure
Piece;HSV assignment of the space reflection device 203 according to the target image of acquisition under the hsv color space, determines the target
The mapping relations of whole pixels and the multiple color sub-spaces in image, to determine the color classification of whole pixels;Face
Color determining device 204 determines the color classification of the target image according to the color classifications of whole pixels;Image provides dress
Put 205 and the color classification of the target image and the target image is supplied to the user equipment, as color classification passes through
The form of form, the mode of figure, the form, the form of webpage etc. that target image passes through file.
Skilled artisans will appreciate that above by mobile phone terminal by the face of the target image and the target image
Colour sorting is supplied to the method for the user equipment to be only for example, and other are existing or are likely to occur the target image from now on
It is supplied to the method for the user equipment to be such as applicable to the present invention with the color classification of the target image, should be also included in this
Within invention protection domain, and it is incorporated herein by reference.
Preferably, image providing device 205, according to the color classification of the target image, the target image is provided
To the user equipment.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, image acquiring device 102, using HTTPs agreements, calls network
The API that sheet data storehouse provides, obtains the image of 1000 " Garfield ", and its storage format is jpg picture;Space reflection device
The 203 HSV assignment according to the target image of acquisition under the hsv color space, determine whole pixels in the target image
With the mapping relations of the multiple color sub-spaces, to determine the color classification of whole pixels;Color determining device 204
According to the color classification of whole pixels, the color classification of the target image is determined;Image providing device 205 is according to the mesh
The color classification of logo image, the target image is supplied to the user equipment, as target image generates according to color classification
Different files, target image are presented etc. according to color classification on the diverse location of webpage.
Skilled artisans will appreciate that above by mobile phone terminal according to the color classification of the target image, by described in
Target image is supplied to the method for the user equipment to be only for example, and other are existing or are likely to occur from now on according to the target
The color classification of image, it is supplied to the user equipment method to be such as applicable to the present invention target image, should also includes
Within the scope of the present invention, and it is incorporated herein by reference.
Preferably, sorting device 1 also includes demand acquisition device 206 and image matching apparatus 207, wherein, demand obtains
Device 206 obtains color demand of the user corresponding to the search sequence to the target image by semantic analysis
Information;Color classification and the color demand information of the image matching apparatus 207 according to the target image, from the target figure
The matching image for selecting the color classification to match with the color demand information as in;Image providing device 205 is by described in
Matching image is supplied to the user equipment.
Specifically, demand acquisition device 206 obtains the user couple corresponding to the search sequence by semantic analysis
The color demand information of the target image.Wherein, semantic analysis including the use of statistics calculate method to substantial amounts of text set
Analyzed, so as to extract potential semantic structure between word and word, and with this potential semantic structure, looked into described in acquisition
Ask color demand information of the user corresponding to sequence to the target image.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, demand acquisition device 206 obtains user by semantic analysis and inquired about
Demand is " Garfield ", and color is " blueness ".
Skilled artisans will appreciate that it is right to be obtained by semantic analysis for the search sequence institute above by mobile phone terminal
The user answered is only for example to the method for the color demand information of the target image, and other are existing or may go out from now on
Now by semantic analysis, the user corresponding to the search sequence is obtained to the color demand information of the target image
Method is such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
Color classification and the color demand information of the image matching apparatus 207 according to the target image, from the mesh
The matching image for selecting the color classification to match with the color demand information in logo image.Such as user is whole by mobile phone
End, internet is connected to by WiFi modes, and it is " blue that text query information is inputted in the cell-phone customer terminal of network picture database
Color Garfield ", it is Garfield that demand acquisition device 206 obtains user's query demand by semantic analysis, and color is blue, image
Acquisition device 102, using HTTPs agreements, the API for calling network picture database to provide, the figure of acquisition 1000 " Garfield "
Picture, its storage format are jpg picture;Space reflection device 203 is according to the target image of acquisition under the hsv color space
HSV assignment, the mapping relations of whole pixels and the multiple color sub-spaces in the target image are determined, with described in determination
The color classification of whole pixels;Color determining device 204 determines the target image according to the color classification of whole pixels
Color classification;Image matching apparatus 207 is according to the color classification of the target image and the color demand information, from described
The Garfield image for selecting the color classification to match with blueness in target image.
Skilled artisans will appreciate that above by mobile phone terminal according to the color classification of the target image with it is described
Color demand information, the matching figure for selecting the color classification to match with the color demand information from the target image
The method of picture is only for example, and other are existing or are likely to occur the color classification according to the target image and the color from now on
Demand information, select that the color classification and the color demand information match from the target image matches image
Method is such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
The matching image is supplied to the user equipment by image providing device 205.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, demand acquisition device 206 obtains user by semantic analysis and inquired about
Demand is Garfield, and color is blueness, image acquiring device 102, using HTTPs agreements, calls network picture database to provide
API, obtain 1000 " Garfield " image, its storage format be jpg picture;Space reflection device 203 is according to acquisition
HSV assignment of the target image under the hsv color space, determine whole pixels and the multiple face in the target image
The mapping relations in dice space, to determine the color classification of whole pixels;Color determining device 204 is according to whole pictures
The color classification of element, determine the color classification of the target image;Image matching apparatus 207 is according to the color of the target image
Classification and the color demand information, the Garfield that the color classification is selected from the target image with blueness to match
The matching image is supplied to the user equipment by image, image providing device 205.
Skilled artisans will appreciate that the matching image is supplied to the user equipment above by mobile phone terminal
Method be only for example, other are existing or are likely to occur the matching image method that is supplied to the user equipment from now on
The present invention is such as applicable to, should be also included within the scope of the present invention, and be incorporated herein by reference.
Preferably, the above-mentioned equipment 1 for being used to carry out target image color classification can be combined with existing browser,
A kind of new browser is formed, existing browser can be the IE browser of such as Microsoft Corporation, Netscape companies
Netscape browser etc..
It is highly preferred that the above-mentioned equipment 1 for being used to carry out target image color classification can also be with existing search engine
It is combined, forms a kind of new search engine.
Fig. 3 show according to a further aspect of the present invention by it is computer implemented be used for target image carry out color classification
Method flow diagram.Here, realize that the sorting device 1 of this method includes but is not limited to individual host, minicomputer, large scale computer, more
Individual host set, network host, single network server, multiple webserver collection, and the cloud that multiple servers are formed.
This, cloud is made up of a large amount of computers or the webserver based on cloud computing (Cloud Computing), wherein, cloud computing is
One kind of Distributed Calculation, a super virtual computer being made up of the computer collection of a group loose couplings.
Specifically, in step S301, sorting device 1 carries out Subspace partition processing to hsv color space, more to obtain
Individual color sub-spaces, wherein, each color sub-spaces correspond to a kind of color.
Wherein, HSV (Hue, Saturation, the Value) color space is three base attributes according to color:Color
Phase (Hue), saturation degree (Saturation) and lightness (Value) determine a kind of method of color.The model in hsv color space
Corresponding to a conical subset in cylindrical-coordinate system, the suitable form and aspect of rounded bottom surface of cone, saturation degree is then from the center of circle to side
Edge increase, lightness are then successively decreased from base to the vertex of a cone, and H, S, V span is respectively 0≤V≤1,0≤S≤1,0≤H < 360.
The top surface of circular cone corresponds to V=1, represents that color is brighter, and color H around the anglec of rotation of V axles by giving.Red corresponds to 0 ° of angle,
Green corresponds to 120 ° of angle, and blueness corresponds to 240 ° of angle.In hsv color model, each color and its complementary color phase
Poor 180 °.Saturation degree S values, so the radius of circular cone top surface is 1, the apex of circular cone, V=0, represent black from 0 to 1, circle
S=0, V=1 at the end face center of cone, white is represented, the summit from the end face center of circular cone to circular cone represents brightness gradually dark ash
Color, the i.e. grey with different gray scales.Other conventional color spaces also have RGB color, HSI color spaces, HSL colors
Space, HSB color spaces, YUV color spaces, Lab color spaces, XYZ color space, Ycc color spaces, CMYK color space
Deng.Relative to other color spaces, subjective understanding of the hsv color space closer to people to color, meet visual face
Color continuity, distribution of color is substantially continuous, is easy to the division of subspace.
For example, in hsv color space, in step S301, sorting device 1 is with the space plane vertical with V axles by V axles
Be divided equally into 10 deciles, will the hsv color space be divided into 10 sub-spaces, if H values are 0-360, S values are 0-1, V
The color space that value is 0-0.1 is hsv color space subspace, and the subspace is labeled as HSV [0-360,0-1,0-
0.1].The color sub-spaces include each color spatial dimension corresponding in each color space subspace, such as red
Curve of the color in color space subspace HSV [0-360,0-1,0-0.1] is f1 (H, S) and f2 (H, S), is fallen in curve f1
(H, S) is with the pixel between f2 (H, S) i.e. labeled as red.
Skilled artisans will appreciate that the method for the above-mentioned subspace that 10 are divided into hsv color space is only to lift
Example, other are existing or are likely to occur from now on to the progress Subspace partition processing of hsv color space, empty to obtain multiple colors
Between method be such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
In step s 302, sorting device 1 obtains the target image for intending carrying out color classification.
Wherein, the form of the target image includes but is not limited to:Bmp, jpg, tiff, gif, pcx, tga, exif,
Fpx, svg, psd, cdr, pcd, dxf, ufo, eps, ai, raw.
Wherein, the target image include being stored in local and or network computer or storage device in, wherein network meter
Calculation machine or storage device include cloud computing.
Wherein, the target image include by it is wired with or wirelessly from the computer or storage device
Obtain.
Such as in step s 302, sorting device 1 is by internet, using HTTPs agreements, from network picture database
The inquiry API provided by database, 1000 target images for intending progress color classification, such as " Garfield " image are obtained, its
Storage format is jpg picture.Wherein network picture database includes being stored on special server, and is stored in cloud meter
It is first-class to calculate platform.
Skilled artisans will appreciate that above by internet, using HTTPs agreements, obtained from network picture database
The method that plan carries out the target image of color classification is only for example, and other are existing or are likely to occur from now on for obtaining plan progress
The method of the target image of color classification is such as applicable to the present invention, should also be included within the scope of the present invention, and to draw
It is incorporated herein with mode.
In step S303, HSV assignment of the sorting device 1 according to the target image under the hsv color space, really
Part or all of mapping relations of pixel and the multiple color sub-spaces in the fixed target image, with determine the part or
The color classification of whole pixels.
Wherein described HSV assignment refers to the HSV value corresponding to each pixel in target image, such as the picture in target image
HSV corresponding to plain x is H=0, S=0.5, V=0.05, is designated as HSV [0,0.5,0.05].
Wherein described mapping relations include the HSV value of each pixel in target image, and color space sky
Between divide, determine the subspace belonging to the pixel, then the color sub-spaces with the subspace are mapped.
For example, the HSV [0,0.5,0.05] of the pixel r in target image e, the hsv color space space vertical with V axles
V axles are divided equally into 10 deciles by plane, will the hsv color space be divided into sub-spaces of 10 sub-spaces, wherein one
Subspace belonging to individual is HSV [0-360,0-1,0-0.1], and in step S303, sorting device 1 determines that the pixel r belongs to son
Space HSV [0-360,0-1,0-0.1], sorting device 1 are reflected pixel r HSV value and the color sub-spaces of the subspace
Penetrate, determine that pixel r falls between red curve f1 and f2, and then determine that pixel r belongs to red, in step S301, point
Kind equipment 1 can by the way of the batch processing or distributed treatment mode to target image in partly or entirely pixel carry out
Mapping, to determine the color classification of the partly or entirely pixel.
Skilled artisans will appreciate that the method that the HSV value of the pixel r in the above-mentioned e with target image determines color classification
It is only for example, other are existing or are likely to occur are assigned according to HSV of the target image under the hsv color space from now on
Value, the mapping relations of part or all of pixel and the multiple color sub-spaces in the target image are determined, with described in determination
The method of the color classification of part or all of pixel is such as applicable to the present invention, should also be included within the scope of the present invention,
And it is incorporated herein by reference.
In step s 304, sorting device 1 determines the target figure according to the color classification of the partly or entirely pixel
The color classification of picture.
For example, the color classification of the target image is determined according to the statistical information of color classification, the pixel of target image
The statistical information of color classification is:Red accounts for 60%, scarlet to account for 15%, orange red to account for 10%, and bright red accounts for 2%, and gold accounts for 2%, green
Color accounts for 3%, and deep sky blue to account for 3%, in step s 304, the statistics that sorting device 1 is classified according to the pixel color of target image is believed
Breath, analyzes red relevant colors and accounts for 92%, and the color classification for determining the target image is red
Skilled artisans will appreciate that the above-mentioned statistical information with by color classification determines the face of the target image
The method of colour sorting is only for example, and other are existing or are likely to occur color point according to the partly or entirely pixel from now on
Class, determine that the method for the color classification of the target image is such as applicable to the present invention, should also be included in the scope of the present invention
Within, and be incorporated herein by reference.
Preferably, it is continuous firing between each step, specifically, in step S301, sorting device 1 is persistently to HSV
Color space carries out Subspace partition processing, to obtain multiple color sub-spaces;In step s 302, sorting device 1 persistently obtains
Take the target image for intending carrying out color classification;In step S303, sorting device 1 continues according to the target image described
HSV assignment under hsv color space, determine part or all of pixel and the multiple color sub-spaces in the target image
Mapping relations, to determine the color classification of the partly or entirely pixel;In step s 304, sorting device 1 continues according to institute
The color classification of part or all of pixel is stated, determines the color classification of the target image.It is to continue not between above steps
Disconnected work, here, it will be understood by those skilled in the art that " lasting " refers to above steps respectively according to setting or real-time tune
Whole mode of operation requires that carrying out hsv color space carries out Subspace partition processing, obtains the target figure for intending carrying out color classification
Picture, the color classification for determining the partly or entirely pixel, and determine the color classification of the target image.
Preferably, in step S301, sorting device 1:
- by machine learning, Subspace partition processing is carried out to hsv color space, to obtain the multiple color sky
Between, wherein, each color sub-spaces correspond to a kind of color.
Wherein, the machine learning includes by computer simulation or realized the learning behavior of the mankind, and self-correction, from
It is dynamic to improve.
For example, in step S301, sorting device 1 is carried out at Subspace partition by machine learning to hsv color space
Reason is as follows the step of machine learning to obtain red color subspace:
Step 1:The positive negative sample of red color is first gathered, positive sample is that people's subjectivity is considered that red sample is added to
Positive sample set, negative sample are people's subjectivity not thinks it is that red sample adds negative sample set, and wherein color card is collected
It can be realized by successive ignition, the sample claims marked color card;
Step 2:In hsv color space, V axles are divided equally into 10 deciles with the space plane vertical with V axles, will be described
Hsv color space is divided into 10 sub-spaces;
Step 3:Initial red curve is first set in every sub-spaces, as red curve formula is:F1 (H, S)
=H+S^a1*b1+c1, f2 (H, S)=H+S^a2*b2+c2, wherein a, b, c are parameter, a, b, and c span is respectively:a
(1/,100 50), b (- 360 360), c (- 360 360), the HSV value of target image pixel fall the mark between two curves
For red;
Wherein initial parameter of curve is arranged to f1 (H, S)=H+S^0*0+0=0, f2 (H, S)=H+S^0*0+30=0;
Step 4:Marked color card is classified using original formulation, obtains ratio of the classification correctly with mistake;
Step 5:According to a, b, c spans, formula f1 and f2 are constantly updated according to step-length, often updates a step-length,
Marked color card is classified, correct and classification error a ratio is obtained, such as takes step-length constant for 5, a, b to c,
Obtain once new formula:F (H, S)=H+5=0 and f (H, S)=H+35=0, then using more new formula to marked color
Sample is classified, and obtains ratio of the classification correctly with mistake;
Step 6:After all testing all possible step size combination, one group of optimal parameter, i.e. classification error are obtained
One group of rate minimum;
Step 7:Using one group of optimal parameter, unmarked color card is classified, the color card of misclassification is added
Enter into corresponding marked color card, such as red sample is assigned in negative sample set, then the sample just adds
Into marked positive sample set, it is assigned to if not the color card of red in positive sample set, is then added to negative sample
In this set;
Step 8:According to the sample set of renewal, step 1 is continued executing with, is iterated, until classification accuracy reaches pre-
Determine threshold value, stop iteration, for example accuracy rate threshold value is 95%;
Step 9:Carry out the training of above-mentioned steps 1 to 8 respectively to each color.
It is skilled artisans will appreciate that above-mentioned by machine learning, to be carried out to hsv color space at Subspace partition
Reason, is only for example with obtaining the method for red color subspace, and other are existing or are likely to occur from now on by machine learning,
Subspace partition processing is carried out to hsv color space, the present invention is such as applicable to obtain the method for red color subspace,
It should be included within the scope of the present invention, and be incorporated herein by reference.
It is highly preferred that in step S301, sorting device 1:
- according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space,
To obtain the multiple color sub-spaces corresponding with the color classification number, wherein, each color sub-spaces corresponding one
Kind color.
Wherein, the preset color classification number including the use of this method system or user according to application need or need
Ask, the color classification number pre-set.
Such as user is only interested in scarlet color, preset color classification number is 1, in step S301, sorting device 1
Above-mentioned steps 1 to 8 are performed according to color classification number, to obtain the corresponding color sub-spaces of scarlet color.
Skilled artisans will appreciate that be 1 above by preset color classification number, by machine learning, with obtain with
The method of the corresponding the multiple color sub-spaces of the color classification number is only for example, and other are existing or from now on may
Occur according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space, to obtain
The method for obtaining the multiple color sub-spaces corresponding with the color classification number is such as applicable to the present invention, should also include
Within the scope of the present invention, and it is incorporated herein by reference.
Preferably, in step s 304, sorting device 1:
- according to the color classification and its respective weights of the partly or entirely pixel, weight and determine the target image
Color classification.
For example, in step s 304, sorting device 1 is according to the color classification of each pixel, and the weight of each pixel, weighting
Determine the color classification of the target image.Or for example, the target figure is determined by the statistical information weighting of color classification
The color classification of picture, wherein red weights are 0.1, golden weights are 0.4, and green weights are 0.4, and blue weights are 0.1, target
Image pixel color classification statistical information be:Red accounts for 50%, and gold accounts for 20%, and green accounts for 20%, and blueness accounts for 10%,
In step S304, statistical information and weights that sorting device 1 is classified according to the pixel color of target image, the mesh after weighting is calculated
The statistical information of the pixel of logo image is:Red accounts for 22.7%, and gold accounts for 36.3%, and green accounts for 36.3%, and blueness accounts for 4.5%,
In step s 304, sorting device 1 determines the face of the target image according to the pixels statisticses information of the target image after weighting
Colour sorting is golden+green.
Skilled artisans will appreciate that above-mentioned weight the determination target image to pass through the statistical information of color classification
The method of color classification be only for example, other are existing or are likely to occur color according to the partly or entirely pixel from now on
Classification and its respective weights, weighting determine that the method for the color classification of the target image is such as applicable to the present invention, should also wrapped
It is contained within the scope of the present invention, and is incorporated herein by reference.
It is highly preferred that in step s 304, sorting device 1:
- according to the color classification of the partly or entirely pixel, and with reference to the partly or entirely pixel in the target
The band of position in image, weighting determine the color classification of the target image.
Determined for example, being weighted by the band of position of the statistical information and pixel of color classification in the target image
The color classification of the target image, a width target image are square, and its picture material is an one-by-one inch photograph, and character image mainly collects
In in center section, target image draw is divided into 3 regions, respectively left, middle and right with the straight line perpendicular to base, wherein
Weights of the yellow in left, middle and right are respectively 0.1,0.8,0.1, and weights of the black in left, middle and right are respectively 0.1,0.8,0.1,
Weights of the blueness in left, middle and right are respectively 0.1,0.8,0.1, and the statistical information of the pixel color classification of target image is:Yellow
Account for 30%, black accounts for 30%, and blueness accounts for 40%, and wherein yellow accounts for 10%, 80%, 10% in left, middle and right respectively, black on a left side,
In, right to account for 10%, 80%, 10% respectively, blueness accounts for 40%, 20%, 40% in left, middle and right respectively, in step s 304, point
The statistical information and the weights of present position that kind equipment 1 is classified according to the pixel color of target image, calculate the target after weighting
The statistical information of image pixel is:Yellow accounts for 40.2%, and black accounts for 40.2%, and blueness accounts for 19.6%, in step S301, classification
Equipment 1 determines the color classification of the target image for yellow+black according to the pixels statisticses information of the target image after weighting
Color.
Skilled artisans will appreciate that above by the statistical information and pixel of color classification in the target image
In band of position weighting determine that the method for the color classification of the target image is only for example, other are existing or from now on may
There is the color classification according to the partly or entirely pixel, and with reference to the partly or entirely pixel in the target image
The band of position, weighting determines that the method for the color classification of the target image is such as applicable to the present invention, should also be included in this
Within invention protection domain, and it is incorporated herein by reference.
Preferably, sorting device 1 is additionally included in step S311 (not shown), and in step S311, sorting device 1 is to described
Target image is pre-processed, to obtain the pretreated target image.
Wherein, the pretreatment operation of the step S311 includes following at least any one:
- processing is compressed to the target image;
- color space mapping processing is carried out to the target image, to obtain the target image under hsv color space
HSV assignment;
- intercepting process is carried out to the target image.
Wherein, before the pretreatment is included in the color classification for determining image, the image is pre-processed, such as compression,
Interception or color space conversion etc., according to different application scenarios, certain conversion, such as target figure are carried out to target image
As color space be RGB color, it is necessary to be transformed into HSV image spaces.
Wherein, the compression processing includes pressing target image in the case where not losing excessive visual information
Contracting, as the lossless compression method of image has a Shannon-Fano codings, Huffman codings, the distance of swimming (Run-length) coding,
LZW (Lempel-Ziv-Welch) codings and arithmetic coding etc., the compression method of image have Karhunen-Loeve transformation coding and DCT to compile
Code etc..By compressing processing to reduce the workload that color of image is classified, color classification efficiency is improved.
Wherein, color space mapping processing is included the color space conversion of target image into hsv color space, such as
RGB color is converted into hsv color space.
Wherein, the intercepting process includes subregion or the gold visual zone by intercepting target image, to scheme
Exemplified by the trichotomy of picture, trichotomy by piece image it is vertical, be horizontally divided into three parts, i.e. piece image is divided into equal size 9 big
Small equal part, the joint position of straight line, i.e. middle position, for being laid out important visible elements, the region
Commonly referred to as gold visual zone, the golden section ratio of image is similar with trichotomy, where the joint of the straight line of golden section ratio
Position than trichotomy intersection point closer to picture centre.
For example, the color space mapping that the RGB color of target image is converted into hsv color space is handled, definition
Pixel s in target image is entered as (R, G, B) in rgb space, wherein R respectively, G, B represents the tax of red, green and blue respectively
Value, RGB span have normalized, and are the real number between 0 to 1, if max is R, the maximum in G, B, and min R, G, B
In reckling, in step S311, sorting device 1 is carried out at color space mapping to each pixel of the target image
Reason, the mapping treatment method in RGB color to hsv color space are as follows:
Max=max (R, G, B)
Min=min (R, G, B)
IfR=max, H=(G-B)/(max-min)
IfG=max, H=2+ (B-R)/(max-min)
IfB=max, H=4+ (R-G)/(max-min)
H=H*60
IfH < 0, H=H+360
V=max (R, G, B)
S=(max-min)/max
It is highly preferred that in step S303, sorting device 1:
- HSV the assignment according to the pretreated target image under the hsv color space, determines the pre- place
The mapping relations of part or all of pixel and the multiple color sub-spaces in target image after reason, with determine the part or
The color classification of whole pixels.
For example, in step S311, sorting device 1 is by target image e RGB color by color space mapping
Reason is converted into hsv color space, wherein the HSV [0,0.5,0.05] of the pixel r after processing, hsv color space is with vertical with V axles
Space plane V axles are divided equally into 10 deciles, will the hsv color space be divided into sub-spaces of 10 sub-spaces,
One of them affiliated subspace is HSV [0-360,0-1,0-0.1], and in step S303, sorting device 1 determines the pixel r
Belong to subspace HSV [0-360,0-1,0-0.1], sorting device 1 enters pixel r HSV value and the color space of the subspace
Row mapping, determines that pixel r falls between red curve f1 and f2, and then determines that pixel r belongs to red, in step S303
In, sorting device 1 can by the way of the batch processing or distributed treatment mode to target image in part or all of picture
Element is mapped, to determine the color classification of the partly or entirely pixel.
Skilled artisans will appreciate that mapped above by by target image e RGB color by color space
Processing is converted into hsv color space to determine that the method for the partly or entirely color classification of pixel is only for example, and other are existing
It is having or be likely to occur HSV assignment according to the pretreated target image under the hsv color space from now on, it is determined that
The mapping relations of part or all of pixel and the multiple color sub-spaces in the pretreated target image, to determine
The method for stating the color classification of part or all of pixel is such as applicable to the present invention, also should be included in the scope of the present invention with
It is interior, and be incorporated herein by reference.
Fig. 4 show in accordance with a preferred embodiment of the present invention by it is computer implemented be used for target image carry out color
The method flow diagram of classification, wherein, this method also includes step S405.The preferred embodiment is described referring to Fig. 4:
Specifically, in step S401, sorting device 1 carries out Subspace partition processing to hsv color space, to obtain multiple color
Space, wherein, each color sub-spaces correspond to a kind of color;In step S402, sorting device 1 obtains passes through user with user
The corresponding target image of search sequence of equipment input;In step S403, sorting device 1 is according to the target image
HSV assignment under the hsv color space, determine part or all of pixel and the multiple color in the target image
The mapping relations in space, to determine the color classification of the partly or entirely pixel;In step s 404, the basis of sorting device 1
The color classification of the partly or entirely pixel, determine the color classification of the target image;In step S405, sorting device
The color classification of the target image and the target image is supplied to the user equipment by 1.Wherein, step S401, step
S402, step S403 and step S404 respectively with that step is corresponded to shown in Fig. 3 is same or similar, therefore here is omitted, and passes through
The mode of reference is incorporated herein.Wherein, the user equipment includes PDA, mobile phone terminal, computer, the number with network savvy
Word terminal etc..
Specifically, in step S402, sorting device 1 obtains relative with the search sequence that user is inputted by user equipment
The target image answered.For example, user uses mobile phone terminal, internet is connected to by WiFi modes, in network picture number
According to text query information " blue Garfield " is inputted in the cell-phone customer terminal in storehouse, in step S402, sorting device 1 uses
HTTPs agreements, the API for calling network picture database to provide, obtain the target image related to " blue Garfield ".
It is skilled artisans will appreciate that relative with the search sequence that user is inputted by mobile phone terminal above by obtaining
The method for the target image answered is only for example, other it is existing or be likely to occur from now on acquisition with user pass through user equipment
The method of the corresponding target image of the search sequence of input is such as applicable to the present invention, should also be included in present invention protection
Within scope, and it is incorporated herein by reference.
In step S405, the color classification of the target image and the target image is supplied to institute by sorting device 1
State user equipment.For example, user is connected to internet, in network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in cell-phone customer terminal, in step S402, sorting device 1 uses HTTPs agreements,
The inquiry API for calling network picture database to provide, 1000 target images related to " blue Garfield " are obtained, it is stored
Form is jpg picture;In step S403, sorting device 1 is according to the target image of acquisition under the hsv color space
HSV assignment, the mapping relations of whole pixels and the multiple color sub-spaces in the target image are determined, it is described complete to determine
The color classification of portion's pixel;In step s 404, sorting device 1 determines the mesh according to the color classification of whole pixels
The color classification of logo image;In step S405, sorting device 1 is by the color classification of the target image and the target image
The user equipment is supplied to, as color classification passes through the form of form, the mode of figure, the shape that target image passes through file
Formula, the form of webpage etc..
Skilled artisans will appreciate that above by mobile phone terminal by the face of the target image and the target image
Colour sorting is supplied to the method for the user equipment to be only for example, and other are existing or are likely to occur the target image from now on
It is supplied to the method for the user equipment to be such as applicable to the present invention with the color classification of the target image, should be also included in this
Within invention protection domain, and it is incorporated herein by reference.
Preferably, in step S405, sorting device 1 is according to the color classification of the target image, by the target figure
As being supplied to the user equipment.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, in step S402, sorting device 1 uses HTTPs agreements, calls
The API that network picture database provides, obtains the image of 1000 " Garfield ", and its storage format is jpg picture;In step
In S403, HSV assignment of the sorting device 1 according to the target image of acquisition under the hsv color space, the target figure is determined
The mapping relations of whole pixels and the multiple color sub-spaces as in, to determine the color classification of whole pixels;In step
In rapid S404, sorting device 1 determines the color classification of the target image according to the color classifications of whole pixels;In step
In rapid S405, the target image is supplied to the user to set by sorting device 1 according to the color classification of the target image
It is standby, as target image according to color classification generates different file, target image according to color classification webpage diverse location
Upper presentation etc..
Skilled artisans will appreciate that above by mobile phone terminal according to the color classification of the target image, by described in
Target image is supplied to the method for the user equipment to be only for example, and other are existing or are likely to occur from now on according to the target
The color classification of image, it is supplied to the user equipment method to be such as applicable to the present invention target image, should also includes
Within the scope of the present invention, and it is incorporated herein by reference.
Preferably, this method also includes step S406 and step S407, wherein, in step S406, sorting device 1 passes through
Semantic analysis, obtain color demand information of the user corresponding to the search sequence to the target image;In step
In S407, color classification and the color demand information of the sorting device 1 according to the target image, from the target image
The matching image for selecting the color classification to match with the color demand information;In step S405, sorting device 1 is by institute
State matching image and be supplied to the user equipment.
Specifically, in step S406, sorting device 1 obtains the institute corresponding to the search sequence by semantic analysis
State color demand information of the user to the target image.Wherein, semantic analysis including the use of statistics calculate method to a large amount of
Text set analyzed, so as to extract potential semantic structure between word and word, and with this potential semantic structure, obtain
Take color demand information of the user corresponding to the search sequence to the target image.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, in step S406, sorting device 1 is obtained by semantic analysis and used
Family query demand is Garfield, and color is blueness.
Skilled artisans will appreciate that it is right to be obtained by semantic analysis for the search sequence institute above by mobile phone terminal
The user answered is only for example to the method for the color demand information of the target image, and other are existing or may go out from now on
Now by semantic analysis, the user corresponding to the search sequence is obtained to the color demand information of the target image
Method is such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
In step S 407, color classification and the color demand information of the sorting device 1 according to the target image, from
The matching image for selecting the color classification to match with the color demand information in the target image.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, in step S406, sorting device 1 is obtained by semantic analysis and used
Family query demand is Garfield, and color is blueness, and in step S402, sorting device 1 uses HTTPs agreements, calls network
The API that sheet data storehouse provides, obtains the image of 1000 " Garfield ", and its storage format is jpg picture;In step S403
In, HSV assignment of the sorting device 1 according to the target image of acquisition under the hsv color space, determine in the target image
The mapping relations of whole pixels and the multiple color sub-spaces, to determine the color classification of whole pixels;In step
In S404, sorting device 1 determines the color classification of the target image according to the color classifications of whole pixels;In step
In S407, color classification and the color demand information of the sorting device 1 according to the target image, from the target image
The Garfield image for selecting the color classification to match with blueness.
Skilled artisans will appreciate that above by mobile phone terminal according to the color classification of the target image with it is described
Color demand information, the matching figure for selecting the color classification to match with the color demand information from the target image
The method of picture is only for example, and other are existing or are likely to occur the color classification according to the target image and the color from now on
Demand information, select that the color classification and the color demand information match from the target image matches image
Method is such as applicable to the present invention, should also be included within the scope of the present invention, and be incorporated herein by reference.
In step S405, the matching image is supplied to the user equipment by sorting device 1.
Such as user is connected to internet, in the mobile phone of network picture database by mobile phone terminal by WiFi modes
Text query information " blue Garfield " is inputted in client, in step S406, sorting device 1 is obtained by semantic analysis and used
Family query demand is Garfield, and color is blueness, and in step S402, sorting device 1 uses HTTPs agreements, calls network
The API that sheet data storehouse provides, obtains the image of 1000 " Garfield ", and its storage format is jpg picture;In step S403
In, HSV assignment of the sorting device 1 according to the target image of acquisition under the hsv color space, determine in the target image
The mapping relations of whole pixels and the multiple color sub-spaces, to determine the color classification of whole pixels;In step
In S404, sorting device 1 determines the color classification of the target image according to the color classifications of whole pixels;In step
In S407, color classification and the color demand information of the sorting device 1 according to the target image, from the target image
The Garfield image for selecting the color classification with blueness to match, in step S405, sorting device 1 schemes the matching
As being supplied to the user equipment.
Skilled artisans will appreciate that the matching image is supplied to the user equipment above by mobile phone terminal
Method be only for example, other are existing or are likely to occur the matching image method that is supplied to the user equipment from now on
The present invention is such as applicable to, should be also included within the scope of the present invention, and be incorporated herein by reference.
It is obvious to a person skilled in the art that the invention is not restricted to the details of above-mentioned one exemplary embodiment, Er Qie
In the case of without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, embodiment all should be regarded as exemplary, and be nonrestrictive, the scope of the present invention is by appended power
Profit requires rather than described above limits, it is intended that all in the implication and scope of the equivalency of claim by falling
Change is included in the present invention.Any reference in claim should not be considered as to the involved claim of limitation.This
Outside, it is clear that the word of " comprising " one is not excluded for other units or step, and odd number is not excluded for plural number.That is stated in device claim is multiple
Unit or device can also be realized by a unit or device by software or hardware.The first, the second grade word is used for table
Show title, and be not offered as any specific order.
Claims (18)
1. it is a kind of by the computer implemented method for being used to carry out target image color classification, wherein, this method includes following
Step:
X carries out Subspace partition processing to hsv color space, to obtain multiple color sub-spaces, wherein, each color sub-spaces
A kind of corresponding color;
Wherein, this method also includes:
A obtains the target image for intending carrying out color classification;
HSV assignment of the b according to the target image under the hsv color space, determine part or complete in the target image
The mapping relations of portion's pixel and the multiple color sub-spaces, to determine the color classification of the partly or entirely pixel;
C determines the color classification of the target image according to the color classification of the partly or entirely pixel;Wherein, the step
Rapid c also includes:
- exist according to the color classification and its respective weights of the partly or entirely pixel, and with reference to the partly or entirely pixel
The band of position in the target image, weighting determine the color classification of the target image.
2. according to the method for claim 1, wherein, the step x also includes:
- by machine learning, Subspace partition processing is carried out to hsv color space, to obtain the multiple color sub-spaces, its
In, each color sub-spaces correspond to a kind of color.
3. according to the method for claim 2, wherein, the step x also includes:
- according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space, to obtain
The multiple color sub-spaces corresponding with the color classification number are obtained, wherein, each color sub-spaces correspond to a kind of face
Color.
4. according to the method for claim 1, wherein, this method also includes:
H pre-processes to the target image, to obtain the pretreated target image;
Wherein, the step b also includes:
- HSV the assignment according to the pretreated target image under the hsv color space, after determining the pretreatment
Target image in partly or entirely pixel and the multiple color sub-spaces mapping relations, with determine it is described partly or entirely
The color classification of pixel.
5. according to the method for claim 4, wherein, the pretreatment operation in the step h includes following at least any one:
- processing is compressed to the target image;
- color space mapping processing is carried out to the target image, to obtain the target image under hsv color space
HSV assignment;
- intercepting process is carried out to the target image.
6. method according to any one of claim 1 to 5, wherein, the step a also includes:
- obtain the target image corresponding with the search sequence that user is inputted by user equipment;
Wherein, this method also includes:
The color classification of the target image and the target image is supplied to the user equipment by r.
7. according to the method for claim 6, wherein, the step r also includes:
- according to the color classification of the target image, the target image is supplied to the user equipment.
8. according to the method for claim 6, wherein, this method also includes:
- by semantic analysis, obtain the user corresponding to the search sequence and the color demand of the target image is believed
Breath;
- according to color classification and the color demand information of the target image, the face is selected from the target image
The matching image that colour sorting matches with the color demand information;
Wherein, the step r also includes:
- by it is described matching image be supplied to the user equipment.
9. a kind of sorting device for being used to carry out target image color classification, wherein, the equipment includes:
Space divides device, for carrying out Subspace partition processing to hsv color space, to obtain multiple color sub-spaces, its
In, each color sub-spaces correspond to a kind of color;
Wherein, the equipment also includes:
Image acquiring device, for obtaining the target image for intending carrying out color classification;
Space reflection device, for the HSV assignment according to the target image under the hsv color space, determine the mesh
The mapping relations of part or all of pixel and the multiple color sub-spaces in logo image, to determine the part or all of pixel
Color classification;
Color determining device, for the color classification according to the partly or entirely pixel, determine the color of the target image
Classification;Wherein, the color determining device is additionally operable to:
- exist according to the color classification and its respective weights of the partly or entirely pixel, and with reference to the partly or entirely pixel
The band of position in the target image, weighting determine the color classification of the target image.
10. equipment according to claim 9, wherein, the space division device is additionally operable to:
- by machine learning, Subspace partition processing is carried out to hsv color space, to obtain the multiple color sub-spaces, its
In, each color sub-spaces correspond to a kind of color.
11. equipment according to claim 10, wherein, the space division device is additionally operable to:
- according to preset color classification number, by machine learning, Subspace partition processing is carried out to hsv color space, to obtain
The multiple color sub-spaces corresponding with the color classification number are obtained, wherein, each color sub-spaces correspond to a kind of face
Color.
12. equipment according to claim 9, wherein, the equipment also includes:
Pretreatment unit, for being pre-processed to the target image, to obtain the pretreated target image;
Wherein, the space reflection device is additionally operable to:
- HSV the assignment according to the pretreated target image under the hsv color space, after determining the pretreatment
Target image in partly or entirely pixel and the multiple color sub-spaces mapping relations, with determine it is described partly or entirely
The color classification of pixel.
13. equipment according to claim 12, wherein, the pretreatment operation in the pretreatment unit include it is following at least
Any one:
- processing is compressed to the target image;
- color space mapping processing is carried out to the target image, to obtain the target image under hsv color space
HSV assignment;
- intercepting process is carried out to the target image.
14. the equipment according to any one of claim 9 to 13, wherein, described image acquisition device is additionally operable to:
- obtain the target image corresponding with the search sequence that user is inputted by user equipment;
Wherein, the equipment also includes:
Image providing device, for being supplied to the user to set the color classification of the target image and the target image
It is standby.
15. equipment according to claim 14, wherein, described image provides device and is additionally operable to:
- according to the color classification of the target image, the target image is supplied to the user equipment.
16. equipment according to claim 14, wherein, the equipment also includes:
Demand acquisition device, for by semantic analysis, obtaining the user corresponding to the search sequence to the target
The color demand information of image;
Image matching apparatus, for the color classification according to the target image and the color demand information, from the target
The matching image for selecting the color classification to match with the color demand information in image;
Wherein, described image provides device and is additionally operable to the matching image being supplied to the user equipment.
17. a kind of browser, including:As any one of claim 9 to 16 to target image carry out color classification
Image classification device.
18. a kind of search engine, including:Color classification is carried out to target image as any one of claim 9 to 16
Image classification device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210091353.8A CN103366178B (en) | 2012-03-30 | 2012-03-30 | A kind of method and apparatus for being used to carry out target image color classification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210091353.8A CN103366178B (en) | 2012-03-30 | 2012-03-30 | A kind of method and apparatus for being used to carry out target image color classification |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103366178A CN103366178A (en) | 2013-10-23 |
CN103366178B true CN103366178B (en) | 2017-12-29 |
Family
ID=49367479
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210091353.8A Active CN103366178B (en) | 2012-03-30 | 2012-03-30 | A kind of method and apparatus for being used to carry out target image color classification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103366178B (en) |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105022752B (en) * | 2014-04-29 | 2019-04-05 | 中国电信股份有限公司 | Image search method and device |
CN106296757A (en) * | 2015-06-09 | 2017-01-04 | 中兴通讯股份有限公司 | A kind of image processing method and device |
CN105761283B (en) * | 2016-02-14 | 2018-12-25 | 广州神马移动信息科技有限公司 | A kind of picture key color extraction method and device |
CN107169105A (en) * | 2017-05-17 | 2017-09-15 | 北京品智能量科技有限公司 | Question and answer system and method for vehicle |
US10740647B2 (en) | 2018-03-14 | 2020-08-11 | Adobe Inc. | Detecting objects using a weakly supervised model |
US11468550B2 (en) | 2019-07-22 | 2022-10-11 | Adobe Inc. | Utilizing object attribute detection models to automatically select instances of detected objects in images |
US11107219B2 (en) | 2019-07-22 | 2021-08-31 | Adobe Inc. | Utilizing object attribute detection models to automatically select instances of detected objects in images |
GB2586678B (en) * | 2019-07-22 | 2022-06-22 | Adobe Inc | Utilizing multiple object detection models to automatically select user-requested objects in images |
US11302033B2 (en) | 2019-07-22 | 2022-04-12 | Adobe Inc. | Classifying colors of objects in digital images |
US11631234B2 (en) | 2019-07-22 | 2023-04-18 | Adobe, Inc. | Automatically detecting user-requested objects in images |
US11468110B2 (en) | 2020-02-25 | 2022-10-11 | Adobe Inc. | Utilizing natural language processing and multiple object detection models to automatically select objects in images |
CN111314614B (en) * | 2020-03-11 | 2021-08-13 | 北京字节跳动网络技术有限公司 | Image processing method and device, readable medium and electronic equipment |
US11055566B1 (en) | 2020-03-12 | 2021-07-06 | Adobe Inc. | Utilizing a large-scale object detector to automatically select objects in digital images |
CN112766278B (en) * | 2020-12-25 | 2022-09-13 | 杭州祐全科技发展有限公司 | Kitchen color code identification and early warning method |
US11587234B2 (en) | 2021-01-15 | 2023-02-21 | Adobe Inc. | Generating class-agnostic object masks in digital images |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101231757A (en) * | 2007-12-07 | 2008-07-30 | 北京搜狗科技发展有限公司 | Apparatus and method for analyzing picture dominant hue as well as application in picture searching thereof |
CN101339556A (en) * | 2008-08-18 | 2009-01-07 | 福建四通石材有限公司 | Method for implementing image color similarity contrast by taper cone coordinate |
CN101763429A (en) * | 2010-01-14 | 2010-06-30 | 中山大学 | Image retrieval method based on color and shape features |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4998287B2 (en) * | 2008-01-25 | 2012-08-15 | ソニー株式会社 | Image processing apparatus and method, and program |
-
2012
- 2012-03-30 CN CN201210091353.8A patent/CN103366178B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101231757A (en) * | 2007-12-07 | 2008-07-30 | 北京搜狗科技发展有限公司 | Apparatus and method for analyzing picture dominant hue as well as application in picture searching thereof |
CN101339556A (en) * | 2008-08-18 | 2009-01-07 | 福建四通石材有限公司 | Method for implementing image color similarity contrast by taper cone coordinate |
CN101763429A (en) * | 2010-01-14 | 2010-06-30 | 中山大学 | Image retrieval method based on color and shape features |
Also Published As
Publication number | Publication date |
---|---|
CN103366178A (en) | 2013-10-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103366178B (en) | A kind of method and apparatus for being used to carry out target image color classification | |
CN102819582B (en) | Quick searching method for mass images | |
CN103714349B (en) | Image recognition method based on color and texture features | |
CN101763429B (en) | Image retrieval method based on color and shape features | |
US7925650B2 (en) | Image management methods, image management systems, and articles of manufacture | |
Recky et al. | Windows detection using k-means in cie-lab color space | |
CN103345758B (en) | Jpeg image region duplication based on DCT statistical nature distorts blind checking method | |
JP4366119B2 (en) | Document processing device | |
CN103440348A (en) | Vector-quantization-based overall and local color image searching method | |
RU2581567C2 (en) | System and method for digital image signal compression using true images | |
CN109214298B (en) | Asian female color value scoring model method based on deep convolutional network | |
CN106250431B (en) | A kind of Color Feature Extraction Method and costume retrieval system based on classification clothes | |
CN106483143A (en) | A kind of solar energy photovoltaic panel dust stratification on-Line Monitor Device and its detection method | |
CN107657619A (en) | A kind of low-light (level) Forest fire image dividing method | |
CN109191428A (en) | Full-reference image quality evaluating method based on masking textural characteristics | |
TWI411968B (en) | A method for image characterization and a method for image search | |
CN108921932B (en) | Method for generating multiple reasonable colorings of black and white figure pictures based on convolutional neural network | |
CA2326631A1 (en) | Representative color designating method using reliability | |
CN110188217A (en) | Image duplicate checking method, apparatus, equipment and computer-readable storage media | |
CN110335666A (en) | Medical image appraisal procedure, device, computer equipment and storage medium | |
Reta et al. | Color uniformity descriptor: An efficient contextual color representation for image indexing and retrieval | |
Berens et al. | Log-opponent chromaticity coding of colour space | |
CN109213886A (en) | Image search method and system based on image segmentation and Fuzzy Pattern Recognition | |
CN109064431B (en) | Picture brightness adjusting method, equipment and storage medium thereof | |
CN103049754A (en) | Picture recommendation method and device of social network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |